Any objections to me
fixing all the occurences of Infiniband in gen2/trunk?
Scott
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit
Roland Dreier wrote:
Or Hi Roland, I have problems cloning your git tree, is it an
Or issue on my side?
I was able to reproduce it but I can't explain it. I can't find any
trace of the static_rate branch in my tree on kernel.org. Maybe
mirrors haven't been updated completely?
So i
Here's some feedback on installation, should I file bugs/enhancements in
bugzilla for these?
0) build.sh does not compile Open MPI, forcing me to run install.sh to
compile Open MPI. This makes it harder to set up a build server used to
just compile the code for installation elsewhere.
1) Too
Since there were no relevant replies, I'm going to use option
1 below.
My reasoning: this is a development area. It'll take some time to
get the code to compile and be even somewhat acceptable,
in the meantime, I don't want to clatter the trunk.
When iser target will get to a working state,
Quoting r. Roland Dreier [EMAIL PROTECTED]:
Subject: [openib-general] Re: IPoIB destructor for 2.6.16-stable?
Michael I don't see any way to fix crashes in ipoib in 2.6.16,
Michael then. Do you?
Unfortunately no. If we could get to the bottom of Hal's crash then I
would be fine
Hi Roland!
For libehca (ehca user verbs) I realized that I also need conversion
functions ibv_rate_to_mult() and mult_to_ibv_rate() similar to the ones for
kernel space. I guess they might be used by others as well and did create a
patch for that, see below. Coud you review it? Thanks!
Regards
On Tuesday 11 April 2006 00:54, Woodruff, Robert J wrote:
I am not sure this is a openib issue, but perhaps rather a bug in the
utilities that assumed the size of a MAC address is 8 bytes.
I am not sure if that is the case with this one, but I know it was
the case in some of the user
On Monday 10 April 2006 23:47, James Lentini wrote:
On Fri, 7 Apr 2006, Dotan Barak wrote:
Hi.
I looked at the file: src/userspace/dapl/dapl/openib/dapl_ib_cq.c,
function: dapl_ib_cq_resize:
In this function, when one wants to resize a CQ, the dapl destroys the
old CQ and
On Monday 10 April 2006 23:30, James Lentini wrote:
On Mon, 10 Apr 2006, Dotan Barak wrote:
who is responsible to change this file with valid data?
The system administrator is supposed to edit this file with the
correct values.
for example:
local IPs
local HCAs and
On Tue, 2006-04-11 at 03:42, Michael S. Tsirkin wrote:
Quoting r. Roland Dreier [EMAIL PROTECTED]:
Subject: [openib-general] Re: IPoIB destructor for 2.6.16-stable?
Michael I don't see any way to fix crashes in ipoib in 2.6.16,
Michael then. Do you?
Unfortunately no. If we
Pradeep Satyanarayana wrote:
I had a question related to this. IBED-1.0-rc3- has it gone through
some minimal touch testing on the various platforms listed below, or
has it been simply compiled on the indicated platforms (with the
exceptions noted below). I did not see any such references to
http://openib.org/bugzilla/show_bug.cgi?id=22
[EMAIL PROTECTED] changed:
What|Removed |Added
Status|NEW |RESOLVED
Resolution|
http://openib.org/bugzilla/show_bug.cgi?id=33
--- Additional Comments From [EMAIL PROTECTED] 2006-04-11 04:57 ---
Are you connecting the ports through a switch? If this is the case then I
believe the problem is that you have to different IP subnets which should be on
different
Hi.
I'm using the dtest from the dapl example folder with the following command
line:
./dtest
./dtest -h IP1(IP1 is the IP of the IPoIB I/F in the remote side)
the output of the test is:
server output:
--
1074 CONNECTED!
1074 Send RMR to remote: snd_msg:
http://openib.org/bugzilla/show_bug.cgi?id=33
--- Additional Comments From [EMAIL PROTECTED] 2006-04-11 05:33 ---
One more thing. You can configure all the ipoib interfaces to reside on the same
IP subnet and this should solve the problem. It worked for me.
--- You are
Title: Message
When
tring to build IBED-1.0-rc3 on 2.6.9-34.EL-smp-x86_64
I got the
following error :
In file included from
On Mon, 10 Apr 2006, Scott Weitzenkamp (sweitzen) wrote:
Any objections to me fixing all the occurences of Infiniband in
gen2/trunk?
The only Infiniband spellings I see are in the ehca driver (32 in
comments, 2 in print statements). If you want to change those, you
should send a patch to
On Tue, 11 Apr 2006, Dotan Barak wrote:
Does anyone have an automatic script / tool that searches for all of
the IB devices (and ports for each device) and create a valid
configuration file?
I don't know of one. It would be useful to have.
___
Title: Message
Moshe Kazir wrote:
When tring to build IBED-1.0-rc3 on
2.6.9-34.EL-smp-x86_64
I got the following error :
In file
included from
/var/tmp/IBED/tmp/openib/openib/src/linux-kernel/infiniband/hw/ipath/ipath_cq.c:36:
On Tue, 11 Apr 2006, Dotan Barak wrote:
can anyone help me with this issue?
Can you ^C and kill the server?
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit
On Tuesday 11 April 2006 17:10, James Lentini wrote:
On Tue, 11 Apr 2006, Dotan Barak wrote:
can anyone help me with this issue?
Can you ^C and kill the server?
yes, there isn't any problem in the host.
i can ^C and kill the server and execute it one more time without any problem
The driver allocates SRQ WQEs size with a power-of-2 size both for Tavor and
for memfree. For Tavor, however, the WQE size is required to be only a
multiple of 16, not a power of 2, and the max number of scatter-gather
allowed is reported accordingly by the firmware (and this is the value
On Mon, 2006-04-10 at 17:59, Sean Hefty wrote:
Hal Rosenstock wrote:
Node A sends an RMPP message. This requires normal RMPP processing.
Node A sends an ACK of the final ACK (I'll call ACK2), giving a new
window.
Node B receives ACKs.
Node B sends the response. This requires normal RMPP
On Tue, 11 Apr 2006, Dotan Barak wrote:
On Tuesday 11 April 2006 17:10, James Lentini wrote:
On Tue, 11 Apr 2006, Dotan Barak wrote:
can anyone help me with this issue?
Can you ^C and kill the server?
yes, there isn't any problem in the host.
i can ^C and kill the
http://openib.org/bugzilla/show_bug.cgi?id=33
--- Additional Comments From [EMAIL PROTECTED] 2006-04-11 09:19 ---
I have onfigured all the ipoib interfaces to reside on the same IP subnet. I hit
the problem after running netperf for a while. Then failed. tcpdump showed that
one side
Thank you for the explanations. This was just what I was looking for.
Pradeep
[EMAIL PROTECTED]
Tziporet Koren [EMAIL PROTECTED] wrote on 04/11/2006 03:49:47 AM:
Pradeep Satyanarayana wrote:
I had a question related to this. IBED-1.0-rc3- has it gone through
some minimal touch testing
Dotan I think that maybe the problem is not in the user level
Dotan stack or in the utilities , but in the linux kernel (in the
Dotan module that handles this socket ioctl)
Yes, the problem is in a patch that Red Hat applies to the 2.6.9
kernel in RHEL4. Full details are in the
p5l2:/usr/src/linux-2.6.16/drivers/infiniband# svnversion .
5988
p5l2:~# [86044.767087] Unable to handle kernel paging request for data
at address 0x0068
[86044.767115] Faulting instruction address: 0xd00018fd4b38
[86044.767132] Oops: Kernel access of bad area, sig: 11 [#1]
Hal Rosenstock wrote:
I don't think that can work. If the request and response are RMPP'd, I
think a direction switch is needed so this can't be done.
A direction switch is only needed if we want to follow the DS RMPP protocol.
Why can't both sides just follow the sender-initiated protocol
Roland wrote,
Dotan I think that maybe the problem is not in the user level
Dotan stack or in the utilities , but in the linux kernel (in the
Dotan module that handles this socket ioctl)
Yes, the problem is in a patch that Red Hat applies to the 2.6.9
kernel in RHEL4. Full details
Sean Hefty wrote:
void ipoib_mcast_join_task(void *dev_ptr)
@@ -553,7 +539,8 @@ void ipoib_mcast_join_task(void *dev_ptr
spin_unlock_irq(priv-lock);
}
- if (!test_bit(IPOIB_MCAST_FLAG_ATTACHED, priv-broadcast-flags)) {
+ if (!test_bit(IPOIB_MCAST_FLAG_ATTACHED,
On Tue, 2006-04-11 at 12:38, Sean Hefty wrote:
Hal Rosenstock wrote:
I don't think that can work. If the request and response are RMPP'd, I
think a direction switch is needed so this can't be done.
A direction switch is only needed if we want to follow the DS RMPP protocol.
Why can't
Roland Dreier wrote:
Looks fine but can you redo this on top of the module unload race fix
once we agree on that? I expect the race fix to go into 2.6.17 and
this API change to go into 2.6.18, so the API change needs to apply on
top of the race fix.
Do you want me to continue to hold off
Hello Troy,
did you unload first all OpenIB modules and then the eHCA module
or the other way around?
Can you see any other message (error data) in /var/log/messages?
It looks like you unloaded the module during an interrupt came in.
Can you sent us the steps / commands you've executed when
I had unplugged, then re-plugged the cable, and then ran the following:
rmmod hcad_mod ib_mthca ib_uverbs ib_ipoib ib_sa ib_mad ib_core
Heiko J Schick wrote:
Hello Troy,
did you unload first all OpenIB modules and then the eHCA module
or the other way around?
Can you see any other message
Or So i am downloading from a mirror of kernel.org and it might
Or work with the a non mirror? what would be the url to download
Or it from kernel.org, all the ones I've tried to derive from
master.kernel.org is the main machine. but I'm not sure whether you
can clone directly from
I haven't been watching this thread so I might be missing the point.
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] Behalf Of Hal Rosenstock
Sent: Tuesday, April 11, 2006 12:48 PM
On Tue, 2006-04-11 at 12:38, Sean Hefty wrote:
Hal Rosenstock wrote:
I
Is there a common set available over iWARP / IB ?
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Richard Is there a common set available over iWARP / IB ?
atomics work fine on IB, at least for the mthca driver.
As far as I know, atomic operations are not part of the iWARP spec.
- R.
___
openib-general mailing list
openib-general@openib.org
Roland wrote,
tcpdump will work fine. It doesn't look at the hardware address in
the structure at all.
- R.
I removed the check and indeed tcpdump does now appear to work in
a modified 2.6.9-34EL kernel. I can add this change to my
backport-to-2.6.9 patches for 2.6.9-34EL for people that
Rimmer, Todd wrote:
It is a bad idea to implement a custom double sided approach. This will
suddenly create various compliance and interop issues. For example Windows
Open Fabrics and Linux OpenSM might not interoperate. Not to mention other
OSs (such as Solaris) which have their own IB
Could you plese check in binaries to svn, or let me get them some other
way? Thanks
Scott
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Bob Woodruff
Sent: Tuesday, April 11, 2006 10:30 AM
To: Roland Dreier (rdreier)
Cc: Jerome Taylor;
James Lentini wrote:
It sounds like the disconnect is being lost. Let me see if I can
reproduce this.
Arlin, have you ever seen this?
No. it runs fine on my systems. It looks like the ping pong test on the
server side did not finish. Can Dotan add a -v switch to the dtest to
help
Could you plese check in binaries to svn, or let me get them some other
way? Thanks
Scott
I'll add it to my next set of backport patches and test RPMS and let you
know
when they are checked in, might be a day or 2, since I will need to
regression test after I make new patches.
woody
I've committed this to svn.
- Sean
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
I'll add it to my next set of backport patches and test RPMS
and let you
know
when they are checked in, might be a day or 2, since I will need to
regression test after I make new patches.
For these RPMs, what is something like
kernel-smp-2.6.9-34.OpenIB.6055.trunk.EL.root.x86_64.rpm? Why
Scott For these RPMs, what is something like
Scott kernel-smp-2.6.9-34.OpenIB.6055.trunk.EL.root.x86_64.rpm?
Scott Why would I want to test that instead of the RH 2.6.9-34
Scott kernel?
Because it has things like the tcpdump problem fixed...
Scott For these RPMs, what is something like
Scott kernel-smp-2.6.9-34.OpenIB.6055.trunk.EL.root.x86_64.rpm?
Scott Why would I want to test that instead of the RH 2.6.9-34
Scott kernel?
Because it has things like the tcpdump problem fixed...
Bummer, there's no way to get
Scott Bummer, there's no way to get just a tcpdump binary that
Scott will work with RHEL4 kernel?
You could do it I guess but no one has come up with the patch...
___
openib-general mailing list
openib-general@openib.org
We also need the following changes to IPoIB for this patch to stand
alone from the multicast changes.
- Sean
---
Index: ulp/ipoib/ipoib_multicast.c
===
--- ulp/ipoib/ipoib_multicast.c (revision 6418)
+++ ulp/ipoib/ipoib_multicast.c
Scott wrote,
Bummer, there's no way to get just a tcpdump binary that will work with
RHEL4 kernel?
Scott
I think that net/core/dev.c is part of the core kernel and not a loadable
module. Also, not sure if tcpdump can be modified to work with the kernel
that has the overflow check.
woody
Quoting r. Sean Hefty [EMAIL PROTECTED]:
Subject: Re: [RFC] [PATCH 2/2 v2] ipoib: convert to use new multicast
interface
Sean Hefty wrote:
void ipoib_mcast_join_task(void *dev_ptr)
@@ -553,7 +539,8 @@ void ipoib_mcast_join_task(void *dev_ptr
spin_unlock_irq(priv-lock);
Michael S. Tsirkin wrote:
void ipoib_mcast_join_task(void *dev_ptr)
@@ -553,7 +539,8 @@ void ipoib_mcast_join_task(void *dev_ptr
spin_unlock_irq(priv-lock);
}
- if (!test_bit(IPOIB_MCAST_FLAG_ATTACHED, priv-broadcast-flags)) {
+ if
Scott Roland, can I talk you into doing it?
Not any time soon... I have plenty to do as it is, and I don't have
that much interest in providing a workaround for a Red Hat bug that
isn't present in the standard kernel.
- R.
___
openib-general
I applied this, thanks.
Note that if you depend on this in libehca, then you won't work with
old releases of libibverbs (which are in Fedora Extras and Debian for
example). So you may want to test for the functions in your
configure.in and include a local copy if you don't find them in
Michael Hmm, but this seems like 2.6.17 material. It should have
Michael the same effect with or without multicast group
Michael patch. Right?
I don't think so. With the current code, it shouldn't be possible to
get to that line with a join of the broadcast group pending.
- R.
Quoting r. Roland Dreier [EMAIL PROTECTED]:
Subject: Re: [RFC] [PATCH 2/2 v2] ipoib: convert to use new multicast
interface
Michael Hmm, but this seems like 2.6.17 material. It should have
Michael the same effect with or without multicast group
Michael patch. Right?
I don't
On 4/11/06, Roland Dreier [EMAIL PROTECTED] wrote:
Or So i am downloading from a mirror of kernel.org and it might
Or work with the a non mirror? what would be the url to download
Or it from kernel.org, all the ones I've tried to derive from
master.kernel.org is the main machine.
Michael S. Tsirkin wrote:
Not sure what changed - I thought new multicast has same API.
I'd have to review the code again then - we had lot of subtle bugs in
this area ...
The new multicast module won't go into 2.6.17.
- Sean
___
openib-general
Michael Not sure what changed - I thought new multicast has same
Michael API. I'd have to review the code again then - we had lot
Michael of subtle bugs in this area ...
Yes, that was my though too: why does the new multicast handling end
up scheduling the multicast join task again
Hello,
I have noticed the recent tcpdump issue related with the RHEL4U3
kernel, I have a cluster to install with Infiniband support and
Redhat EL as the distribution.
I am not completely sure exactly what RHEL4U3 comes with in
terms of the Infiniband support, it definitely has parts of the
Scott wrote,
Bummer, there's no way to get just a tcpdump binary that will work with
RHEL4 kernel?
Scott
You can also apply this patch to your RHEL4 kernel source and rebuild the
kernel.
diff -Naurp linux-2.6.9/net/core/dev.c linux-2.6.9-fixups/net/core/dev.c
--- linux-2.6.9/net/core/dev.c
It looks like some mirrors have synched up. I was just able to clone by
http:// from a mirror with no problem.
- R.
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please
Roger wrote,
Does anyone have any suggestions on exactly how to proceed, here
are the options as I see it:
Use what RHEL4U3 has.
Replace RHEL4U3 kernel with newest released kernel.org kernel and
compile the OpenIB userspace tools for it.
I have tested what is on the RedHat EL4.0 U3 with
Do we have any CentOS 4.2 or 4.3 machines?
Regards,
Robert.
--
Robert Walsh Email: [EMAIL PROTECTED]
PathScale, Inc. Phone: +1 650 934 8117
2071 Stierlin Court, Suite 200 Fax: +1 650 428 1969
Mountain View, CA 94043.
Would anyone be interested in binary RPM's for sles10 ia64 for RC1 ?
I am now working on rc2 RPMS and performance baselines.
I have the following RPM's tested on SGI Altix 3000 3000Bx2 A350 and A330 :
libibat-0.9.0-1.sles10.ia64.rpm
libibat-devel-0.9.0-1.sles10.ia64.rpm
The MVAPICH team is pleased to announce the availability of MVAPICH2
0.9.3-rc0 with the following new features:
- Multi-threading support: This support is available for Gen2, VAPI
and uDAPL transport interfaces. In addition, multi-threading
support for TCP/IP interface (provided by MPICH2
On Tue, 2006-04-11 at 15:31 -0700, Robert Walsh wrote:
Do we have any CentOS 4.2 or 4.3 machines?
People on openib-general can ignore this :-) I meant to send it to an
internal alias. Sorry for the bother.
Regards,
Robert.
--
Robert Walsh Email: [EMAIL
Hi Roland,
Sorry to take this long to response. Thanks for all the enhancements.
I cced some Engenio's engineer who can help to send latest FW to you.
This mostly works for me, but I still see one weird problem. If I
make an FMR to cover IO of size more than 58 * 4096 bytes, the IO
never
Hallo ich bin Irina Nikolaeva und ich vertrete die Firma Btrus .
Unsere Firma Btrus bietet folgende Dienste an:
-Reservierung von teueren Hotels
-Vermietung von Autos
-Kaufen von Flugtickets
Mit der Steigerung der Kundenanzahl, wuchsen auch Ihre Bedurfnisse.Wir haben
noch keine Partner in
Vu Hi Roland, Sorry to take this long to response. Thanks for all
Vu the enhancements. I cced some Engenio's engineer who can help
Vu to send latest FW to you.
Thanks... I haven't been good about following up with Engenio about
this issue (IOs with a single direct region of 58 *
Bob,
I have tested what is on the RedHat
EL4.0 U3 with Intel MPI and it
worked ok, so RedHat EL4.0 U3 has all of the userspace libraries needed
to run MVAPICH, although I have not tried it, but I suspect it will
work.
There is one issue that I ran into with the stock RedHat EL4 U3 release
Use what RHEL4U3 has.
Replace RHEL4U3 kernel with newest released kernel.org kernel and
compile the OpenIB userspace tools for it.
A third option is to keep the RHEL4U3 kernel, and use the OpenIB code
from IBED 1.0 rc3.
Scott
___
LionCub worked fine
with rc2, but neither rc2 nor rc3 SDP is working with Cheetah HCA. netperf
and netserver just hang, then continue when I try to attach with them using
strace.
Anyone have this
combo working?
Scott
___
openib-general mailing
74 matches
Mail list logo